SEASONAL VARIATION OF TURKEY: A CONSENSUS CLUSTERING APPROACH

A “consensus clustering” methodology is applied to long-term (1950-2010) Turkey's meteorological data (temperature, humidity and precipitation) in order to analyze the seasonal variations. The consensus clustering analysis applied is based on the methodology of disturbing the original data using resampling techniques, proposed by Monti et al [2]. We employed four different clustering algorithms (Agglomerative Nesting, Divisive Analysis, Partitioning Around Medoids and K-means) in order to form a consensus solution among different algorithms. Results indicated that Turkey is experiencing longer winter and summer, and rather short spring and fall seasons than usual.